Background: Cerebral amyloid angiopathy (CAA) is a heterogeneous small vessel disease that can occur independently or alongside Alzheimer disease (AD). CAA is diagnosed using the Boston Criteria 2.0, integrating clinical and neuroimaging features, whereas the Cerebrospinal Fluid (CSF) role in clinical practice remains under investigation. This study explores whether CSF biomarkers can identify distinct disease phenotypes, supporting hemorrhagic risk stratification. Methods: We enrolled probable patients with CAA retrospectively (Boston Criteria 2.0) from 2 institutions, collecting clinical, neuroimaging, and follow‐up data alongside core CSF biomarkers (Aβ40 [amyloid β 1–40], Aβ42 [amyloid β 1–42], p‐Tau181 [phosphorylated Tau], total‐Tau). Patients with CAA were stratified applying the Amyloid Tau Neurodegeneration (ATN) research framework, according to the presence of CSF amyloidosis (A + CAA versus A − CAA) and tauopathy (A + T + CAA versus A + T − CAA), and using unsupervised clustering, which defined CAA subgroups based on CSF biomarker levels only. Kaplan‐Meier and Cox regression analyses assessed the predictive value of CSF‐based subgroups for symptomatic hemorrhages during follow‐up. Results: Seventy‐one probable CAA patients (aged 71.77±8.45 years, 66% men, median follow‐up 1.15 years [0.50–2.44]) were enrolled. A + CAA showed a higher prevalence of cortical superficial siderosis than A − CAA (67% versus 25%, P =0.016). A + T − CAA had a greater hemorrhagic risk than A + T + CAA during follow‐up (29 versus 7 events per 100 patient‐years, P =0.010; log‐rank test: P =0.013). Unsupervised clustering identified 2 subgroups, which we defined as pure CAA and CAA‐ADA, with pure CAA presenting more symptomatic hemorrhages during follow‐up (22 versus 0 events per 100 patient‐years, P =0.017; log‐rank test, P =0.011). Conclusions: CSF‐based profiling effectively stratifies CAA phenotypes, offering a promising prognostic tool alongside neuroimaging markers. Further validation is needed to confirm its role in identifying patients with CAA with different natural histories.
1 January 2023 in Advances in Computational Science and Engineering
LMLargo Rosanna Benzi 10 Irccs Ospedale Policlinico San Martino
Largo Rosanna Benzi 10 Irccs Ospedale Policlinico San Martino
+4
Radiomics features extracted from medical images have been shown to correlate with tumor histological biomarkers and patient clinical information. An accurate selection of reliable features is crucial for an efficient prediction/stratification analysis in clinical applications. This work introduces a computational method for evaluating the reliability of radiomics features with respect to image segmentation. To do so, we define four reliability scores that link segmentation variability to feature quality, consistency, robustness and instability. These scores can be used to establish a ranking that helps identifying the best features to employ for the application at hand, allowing an a-priori evaluation of radiomics reliability in clinical applications. We show the performance of the method with Magnetic Resonance images of meningioma patients. In this case, we identified three main groups of features displaying similar behaviors with respect to image segmentation variability.
PurposeMany studies caution against using radiomic features that are sensitive to contouring variability in predictive models for disease stratification. Consequently, metrics such as the intraclass correlation coefficient (ICC) are recommended to guide feature selection based on stability. However, the direct impact of segmentation variability on the performance of predictive models remains underexplored. We examine how segmentation variability affects both feature stability and predictive performance in the radiomics-based classification of triple-negative breast cancer (TNBC) using breast magnetic resonance imaging.ApproachWe analyzed 244 images from the Duke dataset, introducing segmentation variability through controlled modifications of manual segmentations. For each segmentation mask, explainable radiomic features were selected using Shapley Additive exPlanations and used to train logistic regression models. Feature stability across segmentations was assessed via ICC, Pearson’s correlation, and reliability scores quantifying the relationship between segmentation variability and feature robustness.ResultsModel performances in predicting TNBC do not exhibit a significant difference across varying segmentations. The most explicative and predictive features exhibit decreasing ICC as segmentation accuracy decreases. However, their predictive power remains intact due to low ICC combined with high Pearson’s correlation. No shared numerical relationship is found between feature stability and segmentation variability among the most predictive features.ConclusionsModerate segmentation variability has a limited impact on model performance. Although incorporating peritumoral information may reduce feature reproducibility, it does not compromise predictive utility. Notably, feature stability is not a strict prerequisite for predictive relevance, highlighting that exclusive reliance on ICC or stability metrics for feature selection may inadvertently discard informative features.
Coronavirus disease 2019 (COVID-19) in hemodialysis patients (HD) is characterized by heterogeneity of clinical presentation and outcomes. To stratify patients, we collected clinical and laboratory data in two cohorts of HD patients at COVID-19 diagnosis and during the following 4 weeks. Baseline and longitudinal values were used to build a linear mixed effect model (LME) and define different clusters. The development of the LME model in the derivation cohort of 17 HD patients (66.7 ± 12.3 years, eight males) allowed the characterization of two clusters (cl1 and cl2). Patients in cl1 presented a prevalence of females, higher lymphocyte count, and lower levels of lactate dehydrogenase, C-reactive protein, and CD8 + T memory stem cells as a possible result of a milder inflammation. Then, this model was tested in an independent validation cohort of 30 HD patients (73.3 ± 16.3 years, 16 males) assigned to cl1 or cl2 (16 and 14 patients, respectively). The cluster comparison confirmed that cl1 presented a milder form of COVID-19 associated with reduced disease activity, hospitalization, mortality rate, and oxygen requirement. Clustering analysis on longitudinal data allowed patient stratification and identification of the patients at high risk of complications. This strategy could be suitable in different clinical settings.
In this paper we introduce a novel computational procedure to quantitatively investigate how image segmentation affects radiomics feature computation. Specifically, this study introduces four correlation coefficients that quantitatively assess the features' reliability in terms of quality, consistency, robustness, and instability of the features themselves. We validate our analysis in the case of an MRI-based study involving meningioma patients. The proposed approach has been intrinsically conceived for automated radiomics analysis and it is of potential interest for other imaging-driven applications in oncology.